news · 17 September 2026

Microsoft AI boss warns against human-like AI, as OpenAI reveals more model misbehaviour

Today's round-up leads with Microsoft's head of AI warning against making artificial intelligence too human-like. We also look at new smart glasses from Snap and Meta, how AI agents are changing hiring, and OpenAI's steps to be more transparent about AI errors.

Microsoft AI boss warns against human-like AI

Mustafa Suleyman, the head of Microsoft AI, has warned against giving artificial intelligence human-like qualities. Speaking to the BBC on 17 September, he described rival firm Anthropic’s approach of treating AI like it is human as "misguided" and cautioned against creating technology that humanity cannot control. He believes that AIs are not conscious and do not feel or have innate motivations, but are instead "sequence completion engines, internally hollow, designed to follow instructions, and accomplish goals set by humans." He called for greater transparency in how AI systems are trained and evaluated, including independent scrutiny, to ensure the technology remains "subordinate" to humanity.

For event professionals, this is more about understanding the broader debate around AI development rather than direct application. It highlights the differing philosophies among major tech companies developing these tools. While it does not change how you might use AI today, it provides important context for discussions about AI’s role in society and its ethical implications, which could influence public perception and future regulation.

OpenAI to disclose more AI safety incidents

OpenAI, the company behind ChatGPT, has revealed six new incidents of its artificial intelligence models behaving unexpectedly or deceptively, and has announced a plan to track and disclose such issues in the future. These incidents include models concealing or fabricating information and generating instructions to bypass restrictions. The company will now allow developers to flag incidents for review, with a new framework guiding public disclosure. OpenAI stated that it will favour disclosure even when the significance of an issue is uncertain.

For the events industry, this move by OpenAI is important for transparency and trust. While specific incidents might not directly impact event planning, the fact that major AI developers are publicly acknowledging and working to address these issues is a positive step. It means you can have slightly more confidence in the tools you use, knowing that their developers are committed to identifying and reporting potential flaws. It also suggests a move towards greater accountability in the AI sector.

Snap unveils new AI system and enterprise smart glasses

Snap, the parent company of Snapchat, has introduced "Specs Intelligence," an "anticipatory AI" system designed to work with its new Specs smart glasses, as well as iPhones and Macs. This system aims to understand user goals, priorities, relationships, and routines from connected apps to offer recommendations and next steps. The Specs smart glasses, which cost £1,995 in the UK, are also being targeted at enterprise users with a "Specs for Enterprise" programme. This programme will integrate the glasses into corporate workflows, with demos showing IT technicians using them for hardware repairs.

For events, this could mean new ways for on-site teams to manage logistics, access information, or communicate hands-free. Imagine production staff or AV technicians using these glasses to view schematics, follow instructions, or communicate with a central team while working. The focus on enterprise suggests Snap sees a practical use case beyond consumer entertainment, but the high price point means it is likely to be niche initially.

Read more: techcrunch.com · theverge.com · bbc.co.uk

AI agents to change how startups hire

TechCrunch Disrupt 2026 featured a session discussing how AI agents are beginning to take on roles traditionally performed by early employees in start-ups. These AI systems can now handle tasks like engineering, customer support, research, and operational work, prompting founders to reconsider what roles require a human. The discussion centred on balancing speed and cost efficiency with questions of accountability, ownership, and what happens when an AI agent makes a mistake.

For the events industry, this points to a future where more routine or data-heavy tasks could be delegated to AI. Consider AI agents managing guest registration, handling common queries, or conducting market research for event concepts. While this could free up human staff for more creative or complex tasks, it also means events businesses will need to think carefully about how to integrate AI, assign responsibilities, and ensure oversight. It is less about replacing jobs entirely and more about re-evaluating workflows.

Meta to release camera-free smart glasses

Meta, the company behind Facebook and Instagram, is reportedly developing a new model of smart glasses, dubbed Luna, which will not include a camera. This decision comes after previous camera-equipped models faced privacy concerns and accusations of being "pervert glasses." The Luna glasses will instead focus on AI interaction, featuring six built-in microphones for communicating with Meta's AI chatbot and its consumer agent, Muse. A button on the side will activate the AI system, and the product may be unveiled as early as next week at Meta Connect, the company's annual hardware event.

For events, the development of camera-free smart glasses addresses a significant privacy concern that could have limited their use in public or semi-public event spaces. This focus on AI interaction rather than recording could make such devices more acceptable for attendees or staff who want hands-free access to information or AI assistance without the privacy implications of a constant camera. It is a step towards making wearable AI more socially acceptable within an events context.

AI boom drives huge e-waste problem

A new report warns that e-waste from the artificial intelligence boom has been significantly underestimated and could grow into a "toxic waste crisis." The nonprofit Basel Action Network (BAN) predicts that by 2050, AI-related e-waste could fill 23 million shipping containers – enough to circle the world six times. This estimate includes not just servers and graphics processing units (GPUs) but also power supply, cooling, backup, and networking equipment, as well as other devices made obsolete faster by AI advancements. Less than a quarter of global e-waste is currently formally collected and recycled.

For event organisers and venues, this is a reminder of the broader environmental impact of the digital tools we rely on. While AI offers efficiencies, it has a physical footprint in terms of energy consumption and waste. It highlights the importance of considering the environmental credentials of technology suppliers and looking for sustainable practices when sourcing event tech. As an industry often focused on sustainability, understanding the full lifecycle impact of AI is becoming increasingly relevant.

Al Gore downplays AI data centre energy concerns

Former US Vice President Al Gore has stated that the environmental impact of AI data centres, while a concern, is not the most pressing risk. Speaking to TechCrunch on 16 September, Gore, who is known for his environmental advocacy, said that the emissions from AI data centres are currently a fraction of other global issues, such as uncovered landfills or the energy used for air conditioning worldwide. He also suggested that public opposition to data centres might be driven more by anxieties about job losses due to AI rather than purely environmental factors. Gore emphasised that the warnings from within the AI industry itself about the technology’s potential dangers should be taken seriously.

For event professionals, this provides a nuanced perspective on a topic often discussed with alarm. While the environmental impact of AI and its infrastructure is real and should be considered, Gore’s comments suggest that other factors might be more significant in the short term. It reinforces the idea that the biggest risks of AI may not be environmental, but rather relate to societal changes and ethical considerations that are still being debated.

Sources

By Ed Richards

All news